A "real time" dataset generator using device sensors and camera
This Android app is stil a prototype, mainly for educational purposes. The aims of the project are:
- to mantain as far as possible the corrispondence between frames and sensor data
- configurable to be lightweight and not memory/battery consuming
- as fast as possible!
- Record frames (video) and sensor data
- Stream both images and sensors to a remote client
- Upload to FTP
- Save locally on the phone
- Rotate sensor axes
- All of the above, together!
Just start the app. To begin a recording session you can either:
- just launch the python client. The app is preset to automatically start streaming
- click the volume button
- tap the screen
- try any other button you may have.... it may work
More options to enjoy in the settings :)
The app uses OpenCV. There are 2 options to build the app
OpenCV will be linked through the OpenCV Manager app. You will be prompted to install it if you don't have it.
If you don´t want to bother installing more stuff, in this case you will need to ship the correct native libraries together with OpenCV module. To do this just uncomment the ndk section in the app build.gradle
ndk {
moduleName "sensor-logger"
abiFilters 'armeabi-v7a', 'arm64-v8a'
}and the sourceSets section in the openCVLibrary module:
sourceSets {
main {
jniLibs.srcDirs "sdk/native/libs"
}
}where the path of jniLibs.srcDirs is where you will need to put the OpenCV static libraries (not included here) of your choice.
A simple python client is provided to capture the streaming of both sensors and images (because with a browser you can only access a video stream). This is actually pretty slow, being implemented with matplolib, but is a good starting point for collecting and make use of "real-time" data within a data science python application
- OpenCV - BSD